Re: ohci scheduling overrun

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Sat, Oct 08, 2011 at 03:03:22PM +0200, Thomas Klausner wrote:
> On Sat, Oct 08, 2011 at 01:21:05AM +0200, Thomas Klausner wrote:
> > When shutting it down with "shutdown -p" to change BIOS settings (to
> > disable USB legacy mode) I just saw:
> 
> With USB legacy mode disabled, the machine rebooted three times today
> already.
> No ddb, just a plain reboot. The log mentions the overruns, of course.
> 
> Is it really safe to make this panic a warning?

probably not, it looks like there's memory corruption.
What the warning says (if I got it right) is that the controller claims to
have completed transfers from/to a descriptor (the controller gives us back the
descriptor's physical address), but this descriptor can't be found in the
list of active descriptors. So it looks like the controller wrote to
memory that has been freed and maybe already reused for something else
(or to some random physical address that was never used for ohci descriptors
in the first place).
this may be caused by a race condition between the ochi controller
walking the descriptors list and the CPU doing changes to the list ...

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--